Arkansas spent millions on verification technology and lost 18,000 people to coverage in ten months. Georgia spent nearly $100 million on systems and enrolled 6,500 people against a 50,000 target. Both states built technical infrastructure. Neither built the complete system that technical infrastructure requires to function.
ARTICLE SERIES:
- MRWR-2A: Verification Systems
- MRWR-2B: Exemption Systems
- MRWR-2C: The Human Layer
The Series 2 trilogy reveals that work requirements implementation requires three distinct but interdependent infrastructures: technical architecture for verification, policy architecture for exemptions, and human architecture for navigation. States that build all three create systems where people can comply. States that build only one or two create systems where compliance becomes structurally difficult regardless of individual effort.
The Recognition Versus Compliance Divide#
The central tension threading through all three articles is whether systems are designed to recognize existing compliance or to police potential non-compliance. This isn’t a technical distinction. It’s a philosophical choice that determines every operational decision.
MRWR-2A presents this starkly through distributed verification networks. A recognition system assumes most people are working and builds infrastructure to capture that work through credentialed employer submission, payroll integration, and automated data matching. Ohio’s approach, which data-matched two-thirds of their population and exempted them from active reporting, represents recognition logic. The system looks for work, finds it, and removes burden.
A compliance system assumes non-compliance until proven otherwise and builds infrastructure to catch failures. Arkansas’s monthly reporting requirement with coverage termination for missed deadlines represents compliance logic. The system assumes people aren’t working unless they actively prove otherwise each month.
MRWR-2B shows how this distinction shapes exemption design. Recognition-based exemption systems use administrative data to automatically identify qualifying conditions. Someone receiving Social Security disability benefits gets automatically exempted without filing paperwork. Someone with documented serious mental illness in Medicaid claims data triggers automatic medical frailty review. The system recognizes exemption eligibility and acts on it.
Compliance-based exemption systems require individuals to initiate exemption requests, gather documentation, navigate bureaucratic processes, and renew exemptions on state-determined schedules. Arkansas required physician attestations submitted through specific forms within specific timeframes. The burden fell on individuals to prove they qualified for accommodations.
MRWR-2C reveals that this philosophical divide determines human infrastructure requirements. Recognition systems need less human intervention because technology identifies most compliance and exemptions automatically. Navigation focuses on edge cases and genuine complexity. Compliance systems need extensive human infrastructure because technology creates barriers that humans must help overcome. Navigation becomes remediation for system-generated obstacles.
The synthesis insight is that states claiming to build “balanced” systems that verify work while protecting vulnerable populations must choose whether verification means recognition or policing. The middle ground collapses under operational pressure. Either technology recognizes existing work and exemptions, reducing human navigation needs, or technology creates documentation barriers that demand extensive human intervention to prevent unjust coverage loss.
The Exemption Accessibility Paradox#
MRWR-2B establishes a paradox that MRWR-2C makes operationally concrete: exemptions designed to protect people often create barriers that the qualifying conditions prevent people from overcoming.
Medical frailty exemptions require documentation from healthcare providers. But the conditions qualifying for medical frailty often make it difficult to maintain stable provider relationships, navigate appointment systems, follow up on paperwork, or sustain the executive function required for multi-step bureaucratic processes. Serious mental illness, substance use disorders, cognitive disabilities, and chronic homelessness all qualify for medical exemptions. They also all impair the capacity to obtain medical exemption documentation.
Caregiver exemptions require proving care responsibilities. But informal caregiving (caring for disabled family members without formal care arrangements) produces minimal documentation. Families managing care through kinship networks rather than formal systems have difficulty proving what they spend their days doing. The populations most likely to provide informal care (lower-income families, immigrant communities, rural populations) face the highest documentation barriers.
MRWR-2C shows how this paradox compounds through human infrastructure gaps. Professional navigators can help with exemption applications if they have adequate capacity, training, and time. But the populations needing exemptions most are exactly the populations that professional services have historically struggled to reach: people experiencing homelessness, people with serious mental illness, people in active addiction, people in rural areas without CBO infrastructure.
The attempted solution in MRWR-2C is layered human infrastructure combining professional navigators, Community Inclusive Social Enterprises (peer support with compensation), and volunteer networks. But this addresses symptoms rather than causes. If exemption processes were designed for accessibility rather than rigor, the human navigation burden would decrease substantially.
The synthesis insight is that exemption design and human infrastructure requirements are inversely related. Accessible exemptions (automated identification, presumptive eligibility, low documentation burden) reduce navigation needs. Rigorous exemptions (high documentation standards, individual initiation required, frequent renewal) increase navigation needs. States claiming they’ll protect vulnerable populations through robust exemptions while building minimal human infrastructure are making incompatible commitments.
The Technology-Human Balance#
All three articles address technology’s role, but MRWR-2C makes explicit what 2A and 2B imply: technology enables coordination and automation, but humans handle complexity, build trust, and navigate edge cases. The balance between technology and human infrastructure determines who the system serves effectively.
MRWR-2A’s verification architecture relies heavily on technology. Distributed submission networks require credentialing systems, audit algorithms, data matching across wage databases, API integration with employers and gig platforms, real-time compliance dashboards, and automated reporting. This technology works well for straightforward cases: W-2 employees with single employers, salaried workers with consistent hours, people with stable housing and reliable internet access.
But technology fails predictably at complexity. Gig workers with income from multiple platforms that don’t integrate with state systems. Seasonal workers with highly variable hour patterns. Domestic violence survivors requiring confidentiality. People with episodic disabilities. Informal sector workers. Cash-based employment. These populations represent 20-30 percent of the expansion adults subject to work requirements. Technology can’t verify their work without human intervention.
MRWR-2B’s exemption systems face similar limits. Automated identification works for people with clear administrative markers in existing databases (Social Security disability recipients, people with Medicaid claims showing exemption-qualifying diagnoses). It fails for people whose qualifying conditions aren’t documented in accessible databases: undiagnosed mental illness, cognitive disabilities without formal testing, informal caregiving arrangements, trauma-related limitations.
The human layer from MRWR-2C doesn’t just supplement technology. It handles the cases where technology architecture assumptions break down. The person whose serious mental illness isn’t diagnosed because they can’t navigate mental health systems. The caregiver whose full-time care responsibilities leave no time for formal employment or exemption documentation. The seasonal worker whose income pattern looks like non-compliance to automated systems.
MRWR-2C estimates that technology can handle perhaps 70-75 percent of cases with minimal human intervention. The remaining 25-30 percent require human judgment, relationship-building, flexible problem-solving, and sustained support. But this 25-30 percent includes the populations most vulnerable to coverage loss, most likely to experience health deterioration, and most expensive when they eventually return to coverage after acute health crises.
The synthesis insight is that states optimizing for technology efficiency are optimizing for average cases while systematically failing complex cases. The 70 percent that technology serves well could largely comply without extensive systems. The 30 percent that technology serves poorly are exactly the populations that work requirements will most impact. The human infrastructure investment should be proportional to population vulnerability, not to ease of automation.
The Temporal Dimension#
The three articles reveal different temporal patterns that interact problematically.
MRWR-2A describes continuous verification. Work hours accumulate throughout the month. Documentation submits whenever it’s available. Compliance calculates monthly but verification happens daily. This creates continuous administrative burden (monthly submission requirements) but also continuous opportunity (any credentialed source can submit anytime).
MRWR-2B describes periodic exemption renewal. Medical conditions don’t change on state schedules, but documentation requirements do. Someone with permanent paralysis faces annual or semi-annual exemption renewal. Someone providing care for a disabled child faces recurring documentation requirements even though care responsibilities don’t end.
MRWR-2C describes episodic human support. Navigation isn’t continuous. It’s intensive during crisis, light-touch during stability, absent during disengagement. But work requirement demands are continuous, and exemption renewals follow state schedules regardless of individual readiness to engage.
These temporal mismatches create predictable failures. Someone loses their job in week three of the month. Continuous verification would catch this immediately. But if they’re too overwhelmed by job loss to seek navigation support, and if the navigator doesn’t have capacity for proactive outreach, they fail compliance that month despite qualifying for unemployment exemption.
Someone’s exemption expires during a mental health crisis. The chronic condition hasn’t resolved, but they can’t navigate renewal paperwork while symptomatic. By the time they stabilize and can engage with navigation support, they’ve lost coverage. Months later, they return sicker and more expensive.
The synthesis insight is that systems requiring continuous compliance but providing episodic support create gaps where coverage loss becomes inevitable. The temporal architecture of verification and exemption systems must align with the temporal reality of human capacity and navigation availability. States can choose continuous demands with continuous support, or episodic demands with episodic support. Continuous demands with episodic support creates predictable and preventable coverage loss.
The Funding and Scale Reality#
MRWR-2C confronts the fundamental resource constraint that MRWR-2A and 2B mostly bracket. Professional navigation at scale (one CHW per 50-75 people) would require 250,000-370,000 FTEs at $8.75-20 billion annually to serve 18.5 million expansion adults. No state has this funding. The federal administrative match doesn’t cover it. The workforce doesn’t exist. The training infrastructure can’t be built in available time.
This reality forces the layered approach: professional capacity for the most complex 5-10 percent, Community Inclusive Social Enterprises (peers receiving compensation) for the moderate complexity 20-30 percent, volunteer and light-touch support for the remaining 60-70 percent. But even this scaled-back model requires funding that states haven’t allocated.
The resource constraint illuminates choices in MRWR-2A and 2B. States building recognition-based verification systems with accessible exemption processes need less human infrastructure because technology does more work. States building compliance-based verification with rigorous exemption processes need more human infrastructure because technology creates more barriers.
Arkansas built compliance systems with minimal navigation funding. Predictable result: 18,000 coverage losses in ten months, most among people who were working or exempt but couldn’t navigate documentation. Georgia built compliance systems with more navigation funding but still inadequate to need. Result: enrollment far below projections, suggesting many eligible people never successfully navigated application requirements.
The synthesis insight is that technical and policy infrastructure choices in MRWR-2A and 2B have direct fiscal implications for human infrastructure requirements in MRWR-2C. States cannot make technical and policy choices in isolation from resource realities. The system must be designed for affordable implementation, not optimal theoretical function.
Interdependencies and Failure Modes#
The three infrastructures don’t function independently. They create interdependencies that generate specific failure modes when components misalign.
Failure Mode 1: Sophisticated verification technology with inaccessible exemptions and minimal navigation. Technology captures work effectively for straightforward employment. But complex workers can’t comply with rigid verification requirements. They attempt exemptions but can’t navigate documentation requirements. No human infrastructure exists to help. Result: coverage loss among working people with complex employment patterns and vulnerable people who qualify for exemptions.
Failure Mode 2: Accessible exemptions with weak verification technology and minimal navigation. Many people qualify for exemptions and can obtain them relatively easily. But verification technology fails to capture work for people with complex employment. They’re not working enough to verify but not clearly qualifying for exemptions. Without navigation support, they fall through gaps. Result: coverage loss among workers with episodic employment or documentation challenges who don’t neatly fit exemption categories.
Failure Mode 3: Strong verification technology and accessible exemptions but inadequate human infrastructure. Most people comply through automated verification or obtain exemptions through streamlined processes. But the 10-15 percent with genuinely complex situations face systems designed for automation that can’t accommodate their edge cases. Without adequate human infrastructure, they lose coverage despite being working or exempt. Result: relatively good overall retention but systematic exclusion of multiply-burdened populations.
Failure Mode 4: Extensive human infrastructure supporting weak verification technology and inaccessible exemptions. Navigators work heroically to help people comply with systems not designed for accessibility. Individual advocacy prevents some coverage losses. But volume overwhelms capacity. Navigators burn out. People in navigation deserts lose coverage. Result: geographic inequality, navigator exhaustion, and coverage losses concentrated among populations without navigation access.
The synthesis insight is that all three infrastructures must be designed together with explicit attention to interdependencies. Strong performance in one or two domains doesn’t compensate for weakness in the third. The system functions at the level of its weakest infrastructure component.
For Practitioners#
State Medicaid directors face the reality that federal requirements specify verification and exemption frameworks but provide minimal guidance on human infrastructure. The Series 2 trilogy suggests that the human layer isn’t optional or supplementary. It’s the component that determines whether technical and policy infrastructure actually functions. Directors must budget for human infrastructure proportional to technical system complexity and exemption rigor, not proportional to perceived fraud risk.
MCO executives learn that verification technology states build determines care coordination requirements. Plans can advocate for state technology decisions that reduce documentation burden on members and MCOs. But regardless of state choices, MCOs need human infrastructure integrated with care coordination. The member struggling with work verification is probably also struggling with medication adherence, appointment attendance, and chronic disease management. Care coordinators must address all of it simultaneously.
Community organization leaders discover their essential and impossible position. Essential because technical systems cannot handle complexity. Impossible because funding inadequate to need, timelines too short for capacity building, and responsibility for system failures they didn’t cause falls on them. The trilogy suggests CBOs must document burden systematically and advocate for both better system design and adequate navigation funding as linked priorities.
Policymakers confront the reality that verification rigor, exemption accessibility, and human infrastructure investment aren’t independent variables to be set separately. They’re interdependent system components that must be designed together. Policy choices in one domain create resource requirements in others. Rigorous verification creates navigation needs. Inaccessible exemptions create documentation burden. Minimal human infrastructure necessitates automated systems designed for recognition rather than compliance.
What The Next Phase Reveals#
Series 2 establishes infrastructure requirements but leaves implementation pathways underspecified. Subsequent series explore how different stakeholders operationalize these infrastructures. Series 3 examines how MCOs integrate work requirement support into care coordination. Series 4 addresses how semi-annual redetermination creates concentrated pressure on all three infrastructure types simultaneously.
The trilogy demonstrates that work requirements implementation is fundamentally a coordination problem requiring aligned infrastructure across technical, policy, and human domains. States that recognize this and invest accordingly will minimize preventable coverage loss. States that treat it as a pure technology problem or a pure policy problem will watch populations cycle through coverage loss regardless of work status or exemption eligibility.
That outcome is neither philosophically necessary nor operationally inevitable. It’s the predictable result of building incomplete systems and expecting people to navigate gaps that infrastructure design created.
References#
Sommers BD, et al. “Medicaid Work Requirements: Results from the First Year in Arkansas.” New England Journal of Medicine. 2019;381:1073-1082.
Sommers BD, et al. “Consequences of Medicaid Work Requirements in Arkansas: Two-Year Impacts on Coverage, Employment, and Affordability of Care.” Health Affairs. 2020;39(9):1524-1532.
Wagner J, et al. “Pain But No Gain: Arkansas’ Failed Medicaid Work-Reporting Requirements Should Not Be a Model.” Center on Budget and Policy Priorities. August 2023.
Government Accountability Office. “Medicaid Demonstrations: Georgia’s Pathways to Coverage Program Spent Twice as Much on Administrative Costs as on Health Care.” GAO-25-107234. September 2024.
Moynihan D, Herd P, Harvey H. “Administrative Burden: Policymaking by Other Means.” Russell Sage Foundation. 2015.
Lipsky M. “Street-Level Bureaucracy: Dilemmas of the Individual in Public Services.” Russell Sage Foundation. 2010.
Kangovi S, et al. “Effect of Community Health Worker Support on Clinical Outcomes of Low-Income Patients Across Primary Care Facilities: A Randomized Clinical Trial.” JAMA Internal Medicine. 2018;178(12):1635-1643.